Since January 2023, I have served as a Graduate Research Mentor in Rice University's Data to Knowledge (D2K) Lab, leading interdisciplinary student teams in building end-to-end machine learning systems for real-world challenges across healthcare, public safety, industry, and space.
Applied cutting-edge vision-language AI to industrial video data. Built an image captioning pipeline generating natural language descriptions of video frames, designed an embedding-based semantic search for conceptual retrieval beyond keyword matching, and created an enhanced UI enabling safety analysts to query videos in free-form natural language. Advances TechnipFMC's ability to proactively identify safety risks and streamline incident investigations.
Watch demo ↗Transformed video monitoring into an automated pipeline detecting objects of interest in real time and providing operational insights. Used the YouTube-VIS dataset to demonstrate proof-of-concept automated monitoring, with efficient database integration allowing detected objects to be searched and linked back to video for risk management and process optimization.
Watch demo ↗Analyzed epilepsy patient travel patterns to improve healthcare accessibility and optimize LivaNova's resource allocation. Patients often travel long distances for specialized epilepsy care; understanding these patterns is essential for reducing healthcare disparities and improving treatment accessibility through data-driven resource planning.
View demo ↗Identified environmental and demographic factors driving HFD service demands across fire, EMS, hazmat, and motor vehicle emergencies. Results support targeted mitigation strategies, optimized resource allocation, and risk assessments, and lay groundwork for HFD's accreditation with the Center for Public Safety Excellence and potential FEMA grant applications.
Watch demo ↗Extended the Fall 2023 CICU project, refining the ML model predicting length of stay for cardiac ICU patients under 30 days of age and conducting deeper feature importance analysis to surface the most predictive clinical variables for clinical planning and resource allocation.
Watch demo ↗Developed machine learning models to predict the length of stay for cardiac ICU patients, an invaluable tool helping clinicians plan and allocate medical resources. The project explored multiple modeling approaches using multi-source patient data, with findings intended to generalize to hospitals treating diverse patient populations worldwide.
Watch demo ↗Analyzed mental health survey data from tech workers (6 years, OSMI 2016-2021) to identify patterns and predictors of mental health conditions. Built classification models predicting current disorder status, determined factors most affecting mental health over time, and created a public-facing webpage showcasing findings.
Watch demo ↗ GitHub ↗